Mixed-language question answering systems sometimes lack sufficient information to properly answer a question due to data sparsity in the target language. However, the answer may be found in a corpus of a different language, which might be larger or otherwise better suited to the domain. What is needed is a method for both identifying and scoring candidate answers to questions using source documents from other languages.